Determination of optimal measuring sites for fault detection of non-linear systems

Abstract

This paper describes how to determine the optimal set of measuring sites for identifying faults that can be detected if full information about the state variables can be obtained. We present here a new reduced-order time-varying linear observation technique to estimate the state variables in non-linear systems with unknown system parameters. We present a strategy to employ this observation technique to determine an optimal set of measuring sites from among an excess number of sites in the sense of minimizing the measuring cost for fault detection. We also develop a new theoretical result for state observation of non-linear systems with unknown parameters and, for an example application, demonstrate via simulations how to determine the optimal measuring sites and how to detect faults in a non-linear electrical machine.